3-Sigma Rule
When you aggregate prices across venues, you sometimes see numbers that just don’t belong. A single bad feed, a temporary liquidity gap, or a mispriced market can produce a rate that’s wildly different from the rest.
The 3-sigma rule is one way to detect and filter those extremes. It looks at how spread out the values are and marks observations that fall far away from the group.
In simple terms, if most prices cluster together and one price sits way outside that cluster, the rule says “treat that with suspicion.” It’s not perfect, but it’s a common and understandable safety check.
Why It Matters
Outlier filtering can prevent one bad venue from distorting benchmarks and downstream conversions.
Are outliers always “bad data”?
Not always. During fast market moves, an early-moving venue can look like an outlier for a short period. That’s why robust systems combine outlier checks with other signals like liquidity, spreads, and data freshness.
When can the 3-sigma rule fail for market data?
Market data is not always nicely distributed, especially during fast moves. In a real shock, the “outlier” might be the first venue to move rather than a bad print. That’s why many systems apply this rule only when there are enough independent sources and when other quality checks also pass.
Real-World Example
Five exchanges show a BTC/USD benchmark around the same level, but one venue is far away due to thin liquidity. A 3-sigma filter can exclude the odd venue so the benchmark stays stable.
3-Sigma Rule and Exchange Rates API
When building exchange-rate benchmarks from multiple sources, statistical filters can help keep the output reliable. CoinAPI’s Exchange Rates API includes outlier handling as part of maintaining rate quality.